Design of a polarization-insensitive terahertz metamaterial biosensor for glioma tissue identification

Abstract

There is an urgent need to develop label-free, accurate detection techniques for gliomas due to the high aggressiveness and heterogeneity of glioma tissues. In this study, a polarization-insensitive terahertz metamaterial biosensor based on a quadruple rotationally symmetric superunit is proposed to address the issues of insufficient sensitivity and polarization-dependent interference associated with conventional metamaterials in terahertz spectroscopy. The structural parameters were optimized through theoretical modeling and electromagnetic simulations, leading to the design of C4-symmetric metamaterials with stable responses over a wide incidence angle range. These metamaterials effectively mitigate signal distortion caused by the random orientation of metamaterial placement during experiments. The experiments were conducted using a terahertz frequency-domain spectroscopy system (THz-FDS) to detect isolated glioma tissues. The results demonstrate that the sensor maintains polarization insensitivity across the full 0-360° range, significantly enhancing the contrast in dielectric properties between tumor and normal tissues. Furthermore, its resonance frequency shift exhibits a strong correlation with tissue thickness, increasing up to 26 µm, after which it stabilizes and no longer exhibits significant changes. This study confirms that polarization-insensitive metamaterials can overcome existing imaging limitations, reduce operator-induced human errors, and provide a novel, non-invasive detection solution for intraoperative boundary delineation and pathological diagnosis of gliomas, with strong potential for clinical translation.

1. Introduction

Gliomas are the most aggressive primary malignant tumors in the central nervous system, accounting for more than 30% of intracranial tumors, and their high recurrence and low survival rates (especially for glioblastomas, WHO grade IV, with a median patient survival of only 12-15 months) make them a major challenge in the field of neuro-oncology [1–3]. Gliomas of the brain are aggressive and heterogeneous [4], and intraoperative sampling of multiple points is required to determine tumor boundaries. Due to the associated functions of brain tissue, the sampling portion should be as small as possible to minimize the impact of surgery. Therefore, the development of a minimally invasive, highly sensitive, label-free detection technique is crucial for surgical resection and therapeutic evaluation of gliomas. Existing clinical diagnosis mainly relies on imaging techniques such as MRI and CT, which, despite providing morphological information, have limited ability to discriminate the tumor microenvironment, making it difficult to achieve precise differentiation of tumor boundaries [5].

The region of the terahertz (THz) spectrum (0.1-10 THz) is of great technological importance because many biological materials and substances have molecular vibration frequencies in this range [6,7], THz technology is widely used in biomedical applications [8,9]. THz spectroscopy has been reported in the diagnosis of molecular markers of gliomas (e.g., IDH [10], EGFR [11]), but it cannot be judged directly by THz spectroscopy, and it needs to be combined with machine-learning methods. Metamaterials are artificially structured sub-wavelength materials [12], whose extraordinary optical properties can significantly improve the sensitivity of THz detection and further expand the range of practical THz sensing applications [13]. Resonance phenomena, such as Fano resonance [14,15] and toroidal dipole (TD) resonance [16,17], can also be realized by metamaterials with tuned structural parameters. Artificial materials THz sensors can achieve local enhancement and resonance modulation of electromagnetic fields through sub-wavelength structural design, significantly improving detection sensitivity [18]. A bow-tie array THz metamaterial biosensor for trace EGFR detection was reported [19]. C. Weisenstein proposed a biosensor that could be used to distinguish between single-stranded and double-stranded DNA [20]. Metamaterial sensing works on the principle that they generate enhanced local electromagnetic fields during resonant excitation and are sensitive to the dielectric properties of the material near the hypersurface. The enhanced local electromagnetic field can change the resonant frequency and transmission amplitude of the THz wave, thus realizing the high sensitivity characteristics of the sensor [21].

However, single-split ring-based constructions are prone to changes in sensitivity due to strong polarization dependence [22,23]. Meanwhile, in order to obtain a broad-band THz frequency response, most experiments nowadays employ widely used femtosecond laser-pumped THz time-domain spectroscopy (THz-TDS) [24,25] and THz frequency-domain spectroscopy (THz-FDS) [26], which emit linearly polarized THz waves. Rotation of such devices will result in a change in the resonant frequency of the metamaterial, which can adversely affect the performance of the biosensor [25]. The strong dependence of conventional metamaterials on the polarization state of electromagnetic waves severely limits their robustness as a biosensing assay.

To address this challenge, polarization-insensitive metamaterials have become a hot research topic in recent years by introducing high symmetry structures (e.g., cruciform [27], toroidal [28], or quadruple rotational symmetry units [29,30]) to achieve polarization-independent electromagnetic response over a wide range of incidence angles. For example, Liang et al [31] proposed a five-layer microfluidic THz biosensor based on gap-isolated exciton resonance, where the absorption spectrum maintains a high value of more than 99% at a resonance frequency of 0.655 THz when the polarization angle is varied from 0°∼90°, which lays the foundation of a stable detection in complex biological environments. Musa Hamza et al. designed highly sensitive microscale three-band biosensors based on THz MTMs for non-melanoma skin cancer diagnosis [32] as well as biosensors employing dual-band THz metamaterial structures for early cervical cancer tissue detection [33]. In brain glioma detection, such designs can eliminate the interference caused by the randomness of the sensor placement direction during the experiment and improve the accuracy of malignant region identification. In addition, for the analysis of why polarization insensitivity occurs, such studies usually only illustrate the TE mode and TM mode cases [34], but do not analyze the general angle (e.g., 45°) cases, which is a limitation for the illustration of the full range of polarization insensitivity.

In this paper, a polarization-insensitive THz metamaterial based on a fourfold rotationally symmetric superunit is proposed, and its potential application in glioma detection is systematically investigated through theoretical modeling, numerical simulation, and ex vivo tissue experiments. The structure of the full paper is as follows: Part II describes the structural design, electromagnetic properties, and polarization-insensitive mechanism of the metamaterial; Part III details the preparation of glioma samples, the composition of the THz-FDS system, and the experimental procedure; Part IV verifies the polarization-insensitivity of the metamaterials designed in the present study as well as the ability of the metamaterials to discriminate between tumor/normal tissues through experiments; and Part V concludes the advantages of the technology and looks forward to the potential application of the metamaterials in intraoperative rapid diagnosis.

2. Structural design and simulation analysis

2.1. Structure design

To achieve polarization-insensitive effects, this study employs a quadruple rotationally symmetric unit cell structure. Each unit cell consists of four split ring resonator (SRR) elements. Figure 1(a) illustrates a schematic of the metamaterial being irradiated by THz waves polarized along the x-direction and incident normally. Figures 1(b) and (c) show the top view and side view of the unit cell, respectively. The top of the metamaterial features a gold (Au) patterned microstructure layer that supports electromagnetic resonance, while the bottom consists of a quartz substrate layer. The geometric parameters of the metamaterial unit cell are set as follows: period P = 140 µm, small gap a = 20 µm, large gap b = 40 µm, metal line width w = 10 µm, gap width G = 8 µm, split ring side length L = 50 µm, quartz substrate thickness H = 500 µm, and gold structure thickness T = 0.2 µm.

Fig. 1.

Structure and image of the metamaterial. (a) Schematic of THz biosensor; (b) top view and (c) side view of a unit cell. The structural parameters are as follows: P = 140 µm, a = 20 µm, b = 40 µm, w = 10 µm, G = 8 µm, L = 50 µm, H = 500 µm, T = 0.2 µm.

2.2. Software simulation

In this study, numerical simulations are carried out using FDTD software in order to analyze the resonance properties of metamaterials. In the simulation, basic cell boundary conditions in the x and y directions are used to model metamaterials with infinite array sizes in free space with electromagnetic waves propagating along the z-direction in the Ex/Hy mode, i.e., the electric field is in the x-direction and the magnetic field is in the y-direction. The gold SRR layer is modeled as a lossy metal with conductivity σ_Au = 4.561 × 10⁷ S/m, while the quartz backing layer with relative permittivity ε = 3.75 is modeled as a lossy dielectric medium with loss angle tangent of 0.0004. It means that such a lowloss tangent results in almost negligible loss of electromagnetic energy in the dielectric layer. Figure 2(a) illustrates the transmission spectrum of the simulation, which resonates at 0.67 THz. As given in Fig. 2(b), the polarization angle between 0° and 90°, the resonant frequency and amplitude of the sensor peak remain almost constant. Therefore, the metamaterial sensor is polarization insensitive.

Fig. 2.

(a) Transmission transmission spectrum; (b) Transmission spectra at different polarization angles.

2.3. Theoretical analysis of polarization-insensitive

In order to elucidate the principle of polarization insensitivity generation, we choose several typical polarization angles for analysis. Figure 3(a) shows the surface current distribution at different angles; (b) shows the electric field distribution at different angles; and (c) shows the magnetic field distribution at different angles. In Fig. 3(a), the surface currents are distributed along the metal rings of the crystal cell in opposite flow directions and appear in pairs. For the electric field distribution, as shown in Fig. 3(b), pairs of opposite charges can be observed to gather around the splitting gap of each SRR, which suggests that electric dipoles can be excited, and each electric dipole synthesis direction is overdirected by 90° with respect to the azimuthal angle of the electric field incidence (e.g., at a polarization angle of 0 in the direction of the electric field, the direction of the electric dipole synthesis is oriented along the y-axis negative direction). In Fig. 3(c), the magnetic field is induced by the surface current and is symmetrically distributed around the SRR in opposite directions. At incident polarization angles of 0 and 90 degrees, each unit structure forms four closed rectangular rings (red arrows) and generates two pairs of antialigned magnetic dipoles. At incident polarization angles of 45 and 135 degrees, each unit structure forms two closed rectangular rings and produces a pair of anti-aligned magnetic dipoles. Although the number of magnetic dipoles decreases at polarization angles of 45 and 135 degrees compared to polarization angles of 0 and 90 degrees, the intensity is enhanced. It has been reported that the toroidal dipole Fano resonance can be excited by a set of antiparallel magnetic dipoles, and produce a high Q resonance [35,36]. Subsequently, the induced magnetic field rotates around the interstitial gap of the cell to form a closed magnetic field, which generates the toroidal dipole response. The toroidal dipole synthesis direction within the cell structure lags by 90° compared to the incident azimuthal angle (e.g., at a polarization angle of 0 in the direction of the electric field, the toroidal dipole post-synthesis direction is along the y-axis positive direction). In other words, the post-synthesis direction of the toroidal dipole and the post-synthesis direction of the electric dipole follow the same straight line, but in opposite directions.

Fig. 3.

Resonance frequency 0.67 THz at different polarization angles (a) surface current distribution; (b) electric field distribution; (c) magnetic field distribution.

In summary, at incidence at each polarization angle, in this resonance, the magnetic dipoles cancel each other along the positive and negative directions of the z-axis, and the electric and toroidal dipole roles weaken, thus maintaining the polarization insensitivity of the structure.

The above is a qualitative analysis of polarization insensitivity, which is quantitatively analyzed in the following by multipolar decomposition [37,38]. We performed multipolar decomposition calculations in Cartesian coordinate system for the electromagnetic field distribution of the metamaterial at different incident polarization angles, and obtained including Electric Dipole (ED), Electric Quadrupole (EQ), Magnetic Dipole (MD), Magnetic Quadrupole (MQ), and Toroidal Dipole (TD) for the scattered power. Figure (a) shows that the top three contributions to the scattered intensity at the resonance are ED, TD, and MD (shaded in purple), respectively, and that the individual moment components remain constant at different angles. Further, we have also decomposed the top three total scattering contributions of the dipoles (ED, TD, and MD) in the x, y, z direction, as shown in Fig. 4(b)-(d). It can be seen that the component size of each dipole in the x, y direction changes more significantly as the incident polarization angle varies, showing a tendency of increasing and decreasing, but maintaining the total amount unchanged, i.e., the overall polarization insensitivity.

Fig. 4.

(a) Scattering intensity contributions of each component at different angles; (b)-(d) Scattering intensity values of the top three components decomposed in x,y,z directions.

2.4. Analysis of frequency shift versus thickness

Further, in order to explore the detection mechanism of THz metamaterial biosensors, the resonance frequency drift induced by different thicknesses and refractive indices of the measured analytes was simulated using FDTD software. When the refractive index is fixed, the thickness d of the analyte is increased from 0 µm to 50 µm, and the corresponding transmission spectra with thickness are shown in Fig. 5(a). Figure 5(b) shows that a larger frequency shift is induced by an analyte with a larger refractive index for the same thickness. The change in frequency shift is most pronounced when varying the thickness of the same analyte when the thickness is small. As the thickness increases, the increase in the frequency shift slows down and the difference in the frequency shift due to different refractive index analytes increases. This is because the difference in frequency shift due to different refractive indices requires a sufficient optical range to be realized. When the thickness of the analyte exceeds 25 µm, the resonance peak frequency shift of the biosensor tends to saturate, and the frequency shift difference caused by analytes with different refractive indices reaches the maximum. This is due to the skin effect of electromagnetic waves, and the metamaterial resonates only with analytes within a limited thickness range from the surface.

Fig. 5.

(a) Transmission spectra at different thicknesses with fixed refractive index and (b) resonance frequency shift versus thickness for analytes with different refractive indexes

3. Sample preparation and THz experiment

3.1. Metamaterial fabrication

Figure 6 systematically presents the fabrication process of the metamaterial device, which can be divided into six key process links. First of all, the surface purification of quartz substrate is carried out, using a combination of chemical cleaning and thermal drying process: after ultrasonic cleaning with organic solvents to remove surface impurities, it is placed in an oven at a constant temperature of 100 °C for 10 minutes of preheating treatment, which effectively eliminates the adsorbed moisture on the substrate. Subsequently, the film molding process is carried out, using the spin coating process to form a uniform photoresist layer on the surface of the substrate, with the rotational speed set at 3,000 rpm, followed by 120 seconds of pre-drying in a 120°C hot plate to enhance the bonding force of the film layer. The graphic transfer stage employs composite mask technology to generate precise sub-wavelength SRR microstructures on the chromium (Cr) mask by a laser direct writing system. After calibration of the precision optical path, a UV light source is used for contact exposure to achieve pattern selective transfer through the mask. The development process employs an alkaline solution to dissolve the photosensitive material in the unexposed area, followed by a 120°C post-bake curing process to enhance the stability of the pattern structure. The metallization process employs a multi-stage vacuum coating technique, first depositing a 30 nm Cr layer as a transition layer to enhance the structural integrity of the subsequent 200 nm gold film by taking advantage of its excellent interfacial bonding properties. The precious metal layer is deposited by magnetron sputtering to ensure the high purity and uniformity of the conductive layer. Finally, through the organic solvent stripping process, the substrate is immersed in acetone solution for selective etching, effectively removing the redundant metal layer, and ultimately obtaining the structurally complete metamaterial device. The process chain achieves controllable preparation of submicron metal resonance structures through precise control of each process parameter (temperature field, rotational speed, film thickness, etc.), providing standardized samples for subsequent electromagnetic performance testing. The fabrication was performed at Henan Micro & Nano Semiconductor Technology Co., Ltd.

Fig. 6.

Flow chart of metamaterial processing.

3.2. Tissue samples preparation

Tissue samples for this study were provided by Beijing Tiantan Hospital, Capital Medical University. All participating patients gave their informed consent, and the study was approved by the Ethics Committee of Beijing Tiantan Hospital, Capital Medical University. The tumor tissues resected during surgery and the adjacent normal tissues necessary for the procedure were immediately placed in pre-cooled phosphate-buffered saline (PBS) and transported back to the laboratory for further processing. After removing the specimens from the PBS and blotting off surface moisture, the specimens were placed in cryomolds containing frozen tissue embedding medium and stored in a −80°C freezer. The detailed specimen preparation process has been reported in Ref. [39]. Once fully solidified, the tumor and normal tissues were sectioned at varying thicknesses using a cryostat. The sectioning thickness was set to 5, 10, 15, 20, 26, and 30 µm, with three sections at each thickness attached to the surface of the metamaterial. For comparison, an additional three 30 µm sections were placed on a 500 µm thick standard glass slide. Subsequently, THz spectroscopy experiments were conducted.

3.3. THz setup and measurements

In our study, THz-FDS (TeraScan 1550, TOPTICA Photonics, Germany) was used for spectral acquisition, which has been reported in Ref. [40]. The transmitter of the system is a frequency tunable THz source based on nonlinear optical mixing technique. The laser beam is generated by two Distributed Feedback (DFB) semiconductors with a minimum frequency step of 10 MHz. The DFB lasers generate NIR lights with center wavelengths of 1533 nm and 1538 nm, respectively. The frequency difference between the two lasers can be adjusted from 0.05 THz to 1.41 THz. The two laser beams are first mixed and then split. One of the beams radiates onto an optical mixer to produce a THz wave which is focused and carries the sample information to the detector, where it converges with the other beam at the detector. Finally, the two beams generate a photocurrent signal that can be collected by a computer associated with the lock-in amplifier. Four off-axis parabolic mirrors serve to collimate and focus the terahertz waves, and a sample holder with rotating frame is placed horizontally at the focus of the optical path, the structure of which is schematically shown in Fig. 7.

Fig. 7.

Schematic diagram of the THz-FDS setup.

First, the THz spectra of the metamaterials with uncovered tissues were collected and recorded, then the spectra of the metamaterials with different thicknesses of tissues attached were collected, and the data were repeated three times for each experiment to reduce the systematic errors. The data processing algorithm has been described in detail in Ref. [41]. In this study, we denote the resonance frequency of the metamaterial of the uncovered tissue as $f_0$ , and the resonance frequency of the metamaterial of the covered tissue as $f_1$ . The frequency shift $\mathrm{\Delta }f$ is the difference between the two resonance frequencies, i.e. $\mathrm{\Delta }f=f_1-f_0$ .

4. Results and discussion

4.1. Comparison of fabrication test result with simulation result

Figure 8(a) shows the simulated and metamaterial experimental transmission spectra, where the red curve is the simulated curve and the blue is the measured curve of the fabricated metamaterial. The figure shows that the simulated transmission spectrum resonates at 0.67 THz and the fabricated transmission spectrum resonates at 0.66 THz. The experimental curves show periodic slight oscillations on both sides of the resonance frequency, which are caused by THz waves passing through the quartz substrate, and the amplitude of the oscillations is much smaller than the amplitude at the resonance frequency, and thus do not affect the analysis of the resonance. Figures 8(b) and (c) show the design of the unit structure and the machined and fabricated electron micrographs, respectively. Due to the manufacturing technology, there are some errors between the manufacturing and the design, and it can be clearly seen that the opening of the split ring narrows in the fabricated electron microscope diagram. Since the magnitude of the resonance frequency is proportional to the spacing of the split ring openings, this leads to a smaller resonance in the measured results compared to the design results. The measurement results show that although there is a difference in the resonance frequency of the transmission curve, the difference is very small, only 0.01 THz and the envelope of the measurement results is very consistent with the simulation, and in general, the measurement results are consistent with the simulation results. In other words, the fabrication conforms to the design.

Fig. 8.

(a) Simulated and metamaterials experimental transmission spectra; (b) and (c) design drawings of the unit structure and electron microscope images of the machining and fabrication, respectively.

4.2. Polarization-insensitive test

In order to comparatively analyze the advantages of polarization-insensitive metamaterial in practical applications, the polarization-insensitive metamaterial designed in this study and the polarization-sensitive metamaterial mentioned in Ref. [42] were placed in the THz optical path. Their orientations were adjusted by an integrated sample rotation frame to achieve the incidence of THz waves with different angles of polarization directions to the SRR surface. Figures 9(a) and (b) show the transmission spectra curves of the metamaterials without and with 15 µm cancer tissue at different angles, and Fig. 9(c) shows the resonance frequency distributions at different angles, and it can be seen that the frequency magnitude is independent of the polarization angle, which verifies the polarization-insensitive nature of the structure designed in this study. Figure 9(d) demonstrates the frequency shift ranges of the two tests at different phase differences, and it can be seen that the magnitude of the frequency shift is independent of the phase difference of the polarization angle at which the two tests are placed, thus verifying the robustness of the structure designed in this study to the polarization angle. Similarly, Fig. 9(e) demonstrates the resonance frequency distributions of polarization-sensitive metamaterials without and with 15 µm cancer tissue at different angles, and it can be seen that the magnitude of the frequency is related to the polarization angle, and this structure is polarization angle sensitive. Figure 9(f) Demonstrating the frequency shift range for the two tests with different phase differences, it can be seen that the frequency shift magnitude is related to the phase difference of the polarization angle for the two test placements, thus verifying that the placement angle of this structure is more demanding during the test.

Fig. 9.

Polarization sensitivity test of two types of metamaterials at different angles. (a) Transmission spectra curves of bare polarization-insensitive metamaterials at different angles; (b) Transmission spectra curves of polarization-insensitive metamaterials of tumor tissues covered with a thickness of 15 µm at different angles; (c) Polarization diagram representation of polarization-insensitive resonance frequency; (d) Frequency shift range of polarization-insensitive metamaterials for two tests with different phase differences; (e) Polarization diagram representation of polarization-insensitive metamaterials for resonance frequency (f) Frequency shift range of polarization-sensitive metamaterials for two tests with different phase differences.

4.3. Experimental results in glioma tissues

As a comparison, Fig. 10(a) shows the transmission spectra of 5 µm and 30 µm thickness tissues adhered to the metamaterial and 30 µm adhered to ordinary quartz glass. It can be seen that the 30 µm tumor/normal tissue adhered to ordinary glass cannot be seen the difference at all. On the other hand, a significant difference between tumor/normal tissue can be seen at 5 µm thickness on the metamaterial, and the difference between the transmission spectra of the two types of tissues is further increased at 30 µm thickness on the metamaterial. This demonstrates the superior performance of metamaterials in the trace detection of glioma tissues.

Fig. 10.

(a) Comparison of transmission spectra of plain glass and SRR as carriers (b) Frequency shift versus tissue thickness curve.

Figure 10(b) demonstrates the comparison of the magnitude of the frequency shift between tumor tissue and normal tissue at different thicknesses. On the one hand, the curve demonstrates the relationship between the frequency shift and the tissue thickness; at smaller thicknesses, the curve is steeper and the frequency shift size varies significantly with the thickness, but it tends to flatten out after 26 µm and the frequency shift no longer increases with the thickness. This phenomenon is consistent with the performance in simulation, as shown in Fig. 5(b). On the other hand, it can be seen that the normal tissue frequency shift is larger than that of the tumor tissue at each thickness, which gives us the possibility of detecting brain tumors by metamaterials. In our previous study [39], it can be understood that the reason for this difference is that the tumor tissue is necrotic, the organelles are cleaved into lipids, and the cell density is reduced compared to normal tissue, which leads to a smaller refractive index, and therefore shows that the normal tissue frequency shift is larger than the tumor tissue at the same thickness.

4.4. Discussion of clinical applications

Currently, the diagnosis of gliomas primarily relies on obtaining tissue samples through tumor resection or biopsy, followed by histopathological analysis to determine tumor grade and delineate its boundaries. However, conventional methods such as H&E-stained tissue sections involve complex procedures, and intraoperative frozen section analysis is often limited by factors including sampling location, subjective interpretation by pathologists, and variability in tissue preparation. These limitations can affect both the accuracy and sensitivity of diagnosis [43]. In recent years, THz spectroscopy has attracted increasing attention in tumor diagnostics due to its non-invasive nature, label-free detection, and rapid analysis capabilities [44,45]. With the aid of metamaterial-enhanced sensing technologies, even trace amounts of brain tissue are sufficient to distinguish pathological differences, offering new possibilities for more precise intraoperative diagnosis in the future. THz technology holds promise for real-time intraoperative pathological assessment of gliomas, potentially providing surgeons with immediate guidance on tumor margins and assisting in optimizing the extent of resection.

5. Conclusion

In this study, we presented a highly sensitive terahertz metamaterial sensor based on polarization insensitivity. The metamaterial consists of a quartz substrate and a gold structure. Simulations show that the transmittance spectrum of the sensor exhibits a high degree of coherence from 0-90 degrees, and we explore the reason why the metamaterial sensor remains polarization-insensitive through pattern analysis. In addition, we apply it to the detection of glioma tissues, and the experimental results show that the frequency shifts of normal tissues are larger than those caused by tumor tissues at different thicknesses in the range of 5-30 µm, and they are not affected by the polarization angle. The frequency shifts of normal and tumor tissues show a clear difference at a thickness of 5 µm, which indicates its ability to detect small amounts of material. As the thickness increases, the difference in frequency shift caused by different tissues gradually widens, which facilitates the discrimination of diseases. The polarization-insensitive metamaterials designed in this study do not cause bias in the results due to the experimenter's placement of the sensor orientation during the actual detection use, which enhances the robustness of sensing. This is important for disease diagnosis, detection of cancer markers and analysis of biological samples. Therefore, the polarization-insensitive sensors proposed in this study have great prospects for development and application.

Funding

National Natural Science Foundation of China 10.13039/501100001809 ( 62005014, 62205348, 81903060, 81902533); Beijing Municipal Administration of Hospitals 10.13039/501100009601 ( No. PX2024019).

Disclosures

The authors report no competing interests.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Supplemental document

See Supplement 1 ^{(3.4MB, pdf)} for supporting content.